A Target Tracking Optimization Method Based on Tracking Learning Detection

Abstract

The invention discloses a target tracking optimization method based on tracking learning detection, which belongs to the field of computer vision, and its main idea is to determine a sequence of L grayscale image video frames, and each grayscale image video frame sequence contains a tracking target ; The position of the tracking target in the video frame sequence of the first gray-scale image is known, and the position of the tracking target in the remaining L-1 gray-scale image video frame sequence is unknown; t∈{1,2,...,L}, t The initial value is 1; select a uniform tracking point in the target bounding box of the tth frame, and then obtain the tracking phase bounding box of the t+1th grayscale image video frame sequence in the t+1th frame grayscale image video frame sequence tb t+1 ; Obtain the detection stage bounding box db of the t+1th frame grayscale image video frame sequence from the t+1th frame grayscale image video frame sequence t+1 , and then determine the final position of the tracking target in the t+1th frame; add 1 to the value of t until the final position of the tracking target in the second frame to the final position of the tracking target in the Lth frame is obtained, and recorded as the target detected based on tracking learning Track optimization results.

Description

technical field [0001] The invention belongs to the field of computer vision, in particular to a target tracking optimization method based on tracking learning detection, that is, a target tracking optimization method based on tracking-learning-detection (Tracking-Learning-Detection, TLD), which is suitable for video frame sequences Long-term stable tracking of a single target in Background technique [0002] In recent years, object detection and object tracking have been the high-profile scientific and technological frontiers in the field of computer vision, mainly including detecting and identifying interesting moving objects from video frame sequences, and then continuously and stably tracking moving objects, and The motion state of the moving target is described in detail in real time; target detection and target tracking involve disciplines in many fields such as artificial intelligence, image processing, and pattern recognition. [0003] With the rapid development of ...

AI Technical Summary

Problems solved by technology

[0005] The target tracking algorithm TLD can be divided into three stages: the tracking stage, the detection stage and the learning stage. The tracking stage and the detection stage are independent of each other. Tracking ability; the final output of the tracking phase and the detection phase are the bounding box of the tracking phase and the bounding box of the detection phase respectively, and the final output obtained by combining the tracking phase and the detection phase is called the target bounding box; the robustness and stability of the target tracking algorithm TLD The performance is worthy of recognition, but the overall real-time performance of the algorithm is average and needs to be improved, especially in the detection stage of the video frame sequence, the entire gray image video frame sequence must be scanned globally every time, which seriously affects the detection speed of the detection stage , thereby reducing the real-time performance of the entire target tracking algorithm TLD

Images